Phi-3.5-vision-instruct
Microsoft · microsoft/Phi-3.5-vision-instruct
Weights license: MIT · code and components may use other terms
Quality
Quality · LMArena · vision
LMArena 2026-10-19 · human-preference rating · higher is better · this model outlined
- Color = publisher
Source: LMArena 2026-10-19 · collected 2026-10-25 · dataset · CC BY 4.0 · models without a rating are not drawn (no rating ≠ 0)
919 ± 16 · LMArena · vision · Confidence interval 903–935 · 2,592 votes
Reported by the publisher
- Art Style
- 87.2Release Notes
- Counting
- 54.2Release Notes
- Forensic Detection
- 92.4Release Notes
- Functional Correspondence
- 29.2Release Notes
- IQ Test
- 25.3Release Notes
- Jigsaw
- 68.0Release Notes
- Multi-View Reasoning
- 54.1Release Notes
- Object Localization
- 49.2Release Notes
Every reported result
- Relative Depth
- 69.4Release Notes
- Relative Reflectance
- 37.3Release Notes
- Semantic Correspondence
- 36.7Release Notes
- Spatial Relation
- 65.7Release Notes
- Visual Correspondence
- 53.5Release Notes
- Visual Similarity
- 83.0Release Notes
- Overall
- 57.0Release Notes
- short (<2min)
- 60.8Release Notes
- medium (4-15min)
- 47.7Release Notes
- long (30-60min)
- 43.8Release Notes
- Overall
- 50.8Release Notes
- MMMU (val)
- 43.0Benchmarks · Popular aggregated benchmark · 0-shot
- MMBench (dev-en)
- 81.9Benchmarks · Popular aggregated benchmark · 0-shot
- ScienceQA (img-test)
- 91.3Benchmarks · Visual scientific knowledge reasoning · 0-shot
- MathVista (testmini)
- 43.9Benchmarks · Visual math reasoning · 0-shot
- InterGPS (test)
- 36.3Benchmarks · Visual math reasoning · 0-shot
- AI2D (test)
- 78.1Benchmarks · Chart reasoning · 0-shot
- ChartQA (test)
- 81.8Benchmarks · Chart reasoning · 0-shot
- TextVQA (val)
- 72.0Benchmarks · Document Intelligence · 0-shot
- POPE (test)
- 86.1Benchmarks · Object visual presence verification · 0-shot
Values as printed on the card, with the publisher's own settings (card section: Release Notes; Benchmarks); column “Phi-3.5-vision-instruct”. Not comparable across publishers and not used in any ranking. Model card
Quality × memory
No comparable quality × memory pair
This entry has no score on a benchmark that is drawn against planning memory.
Microsoft released Phi-3.5-vision-instruct on 2024-08-20: a vision-language model that answers about images, with 4.1B parameters on the Hugging Face count and the MIT license. Languages the publisher declares: Multilingual.
License: commercial use allowed. At Q4 the weights load about 3.1 GB before each machine's reserve; image resolution, video frames and context add memory, and the vision encoder needs support in the runtime you pick.
Runs on
Fast memory holds it 118
The planning estimate fits the memory the model can use quickly: graphics card VRAM, or unified memory within the macOS GPU share.
Show all 118
Within memory, outside the fast part 1
Within the machine's memory, outside its fast part: on a graphics card the layers that do not fit run on the CPU from PC RAM (the listed RAM, or 32 GB assumed); on a Mac, above the macOS default GPU share. Not a speed figure.
Hardware profiles grouped by the Q4 planning estimate on each machine (dedicated GPUs reserve 2 GB; unified and system memory 25%, 2 to 8 GB): the fast memory holds it, it is within memory but outside the fast part (layers offloaded to PC RAM, or above the macOS GPU share), or the machine is one memory tier short. Not a measured fit or speed.
Measured speed
No published measurement for this checkpoint
Tokens per second are shown only when a source publishes them with hardware, build and setup. They are never estimated.
How to run
Runtime compatibility and tokens per second have not been measured for this entry. Published metadata and a memory estimate do not prove a successful local run.
Library in source: transformers
Documented ways to run it
Copied as printed in the linked source (read 2026-10-18); not run by this site.
from PIL import Image
import requests
from transformers import AutoModelForCausalLM
from transformers import AutoProcessor
model_id = "microsoft/Phi-3.5-vision-instruct"
# Note: set _attn_implementation='eager' if you don't have flash_attn installed
model = AutoModelForCausalLM.from_pretrained(
model_id,
device_map="cuda",
trust_remote_code=True,
torch_dtype="auto",
_attn_implementation='flash_attention_2'
)
# for best performance, use num_crops=4 for multi-frame, num_crops=16 for single-frame.
processor = AutoProcessor.from_pretrained(model_id,
trust_remote_code=True,
num_crops=4
)
images = []
placeholder = ""
# Note: if OOM, you might consider reduce number of frames in this example.
for i in range(1,20):
url = f"https://image.slidesharecdn.com/azureintroduction-191206101932/75/Introduction-to-Microsoft-Azure-Cloud-{i}-2048.jpg"
images.append(Image.open(requests.get(url, stream=True).raw))
placeholder += f"<|image_{i}|>\n"
messages = [
{"role": "user", "content": placeholder+"Summarize the deck of slides."},
]
prompt = processor.tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
inputs = processor(prompt, images, return_tensors="pt").to("cuda:0")
generation_args = {
"max_new_tokens": 1000,
"temperature": 0.0,
"do_sample": False,
}
generate_ids = model.generate(**inputs,
eos_token_id=processor.tokenizer.eos_token_id,
**generation_args
)
# remove input tokens
generate_ids = generate_ids[:, inputs['input_ids'].shape[1]:]
response = processor.batch_decode(generate_ids,
skip_special_tokens=True,
clean_up_tokenization_spaces=False)[0]
print(response)Technical details
What the source and the catalog store for this checkpoint. Missing fields are left out, never zero.
- Checkpoint / source
- microsoft/Phi-3.5-vision-instruct
- Task
- image-text-to-text
- Architecture
- Phi3VForCausalLM
- Library in source
- transformers
- Weight license
- MITmicrosoft/Phi-3.5-vision-instruct model card (Hugging Face) · read 2026-10-18source
- Commercial use
- Allowed MIT license text (SPDX) · read 2026-10-18license text
- Languages
- MultilingualModel card metadata · read 2026-10-25source
- Released
- What's new Phi-3.5 Family - Microsoft PhiCookBook (GitHub, microsoft) · read 2026-10-18source
Every stored field
- Revision checked
- 12b77fb40b63a2c73c68243d3f767aab688a1b2a
- Architecture
- Phi3VForCausalLM
- Library
- transformers
- Hub task tag
- image-text-to-text
- Formats
- transformers, safetensors
- Access
- Open download
- Publisher kind
- Official
- Repository created
- 2024-08-16creation is not the model's release
- Repository updated
- 2025-12-10a commit is not a release date
- Hub downloads (30 days) · likes
- 486,719 · 748this repository only · read 2026-10-25
- Catalog state
- Hub config or weights found; no inventory benchmark link
Sources
Identity and published specifications come from the checkpoint page. Benchmark scores belong to the named release and do not measure speed on your computer.
- Publisher model card
- License: microsoft/Phi-3.5-vision-instruct model card (Hugging Face)
- Release date: What's new Phi-3.5 Family - Microsoft PhiCookBook (GitHub, microsoft)
Curated facts checked: 2026-10-18. The license and release date cite the publisher's sources; parameters, files and task come from the Hugging Face record, and memory is a labelled estimate from its parameter count. A value no source states stays empty; a repository commit is not a model release date.






























































